[Jim Robinson]

"What about 3 or 4 live threads - identified as such? Maybe spread the load?"

Well, we can give it a try. If there’s a chance that hundreds of people trying to post to the same thread at the same time is causing a deadlock on the db, it’s worth a shot to try to split the load. John says the load avg on the web server was low, the cpu’s were not busy and the db sever was mostly idle last time, but we couldn’t get any throughput. Could be a logjam in the db caused by record locks making it hard to get anything in or out.

[Recovering Ex-hippie]

Fr has the same problems that all threads have when something big is going on..it gets slammed.
BTW, do images, pictures have an effect on slowing down a thread or is that irrelevant?

[onyx]

Let’s give it a try.

Start another one, Jim.

Nobody will dare to complain if it’s your duplicate.

Well, on second thought....LOLOLOL.

[ATOMIC_PUNK]

Are you sure they are all live ? lol

[justlurking]

Well, we can give it a try. If there’s a chance that hundreds of people trying to post to the same thread at the same time is causing a deadlock on the db, it’s worth a shot to try to split the load. John says the load avg on the web server was low, the cpu’s were not busy and the db sever was mostly idle last time, but we couldn’t get any throughput. Could be a logjam in the db caused by record locks making it hard to get anything in or out.

Check with John about the DB schema before you try that. Depending on how it is organized, more threads may make it worse.

If it is a lock contention problem, then you need to look deeper into the database performance statistics, and especially the I/O (reads/sec and writes/sec) on the database device.

If it is indeed I/O contention, one thing that would help a great deal is:

[imagine Keanu/Neo's voice]

RAM. Lots of RAM.

If the database isn't using it, the O/S will use it for file caching. Any I/O operation that is avoided will be performed much faster -- as in orders of magnitude.

It isn't unusually for me to see database servers with 64GB of RAM. It makes a huge difference.

[1_Rain_Drop]

Would putting the Freepathon on hold for a short while help a tad?
(I don’t like asking this question)

[PA Engineer]

Well, we can give it a try. If there’s a chance that hundreds of people trying to post to the same thread at the same time is causing a deadlock on the db, it’s worth a shot to try to split the load. John says the load avg on the web server was low, the cpu’s were not busy and the db sever was mostly idle last time, but we couldn’t get any throughput. Could be a logjam in the db caused by record locks making it hard to get anything in or out.

Jim

At first I thought the problem was here. I thought maybe people were refreshing the thread too often. Now I am not so sure.

When posters mentioned twitter, I tried to log on. Overloaded and no response. Then another Freeper mentioned that he did a ping to Freerepublic. The response was over 20 iterations that went through a Chicago server.

Then I tried to go to the Frank Luntz website. Again it was hammered.

I think the problem is elsewhere. Funny how google worked just fine until you hit the threads on any of the trending sites. It struck me as an intentional bottleneck.

A possible suggestion in the future would be to have an autoupdate feature for a selected thread. I am just do not believe the problem is here. I saw it on many other robust sites tonight.

[SoConPubbie]

Well, we can give it a try. If there’s a chance that hundreds of people trying to post to the same thread at the same time is causing a deadlock on the db, it’s worth a shot to try to split the load. John says the load avg on the web server was low, the cpu’s were not busy and the db sever was mostly idle last time, but we couldn’t get any throughput. Could be a logjam in the db caused by record locks making it hard to get anything in or out.

Hey Jim, about the indexing on the database tables.

Sometimes, the simplest things get overlooked.